RFR: 8227168: Cleanup usage of NEW_C_HEAP_ARRAY
kim.barrett at oracle.com
Tue Sep 3 18:07:56 UTC 2019
> On Sep 3, 2019, at 1:19 PM, coleen.phillimore at oracle.com wrote:
> On 9/3/19 12:12 PM, Leo Korinth wrote:
>> I have tried to use FREE_C_HEAP_OBJ correctly, and I did revert one usage of NEW_C_HEAP_OBJ to NEW_C_HEAP_ARRAY as it seemed to be able to also be an array type when allocated in another place.
> […] But I don't agree. I don't think NEW_C_HEAP_ARRAY should be used to allocate and FREE_C_HEAP_OBJ should be used to deallocate. That's worse.
That wasn’t what was suggested. The suggestion was to always be consistent with the pairing,
e.g. use NEW_C_HEAP_ARRAY and FREE_C_HEAP_ARRAY for an object, or use NEW_C_HEAP_OBJ
and FREE_C_HEAP_OBJ, but never mix the pairings for a given object. I think that’s what the
updated change is doing.
> Actually, I think neither NEW_C_HEAP_OBJ or FREE_C_HEAP_OBJ should exist. They should just use new Type(), where Type is inherited from CHeapObj<mtAppropriate>
I was expecting that to have widespread impact, but it seems there are far fewer uses of NEW_C_HEAP_OBJ
than I was expecting. But I suggest that doing anything about that is out of scope for this change.
I’m also not really fond of the allocation base class approach; C++11 allows some alternatives.
More information about the hotspot-dev